On the Feasibility of Learning, Rather than Assuming, Human Biases for Reward Inference.

Rohin Shah, Noah Gundotra, Pieter Abbeel, Anca D. Dragan

Browse the full ICML paper archive.